AI Overviews vs paid search: 3 actions to cut through
Summarize this article with AI
The phone call that changes everything
A client calls me on a Tuesday morning. He has 800 SKUs. 4 000 organic sessions per month. $12 000 in paid budget.
He says: « The ads aren’t clicking anymore. »
I look at his 47 transactional queries. The AI Overview occupies position zero on 31 of them. Paid falls below. Cost per click rose 17%. Ad revenue dropped 23% in six weeks.
The issue is the decision architecture.
He has three options. Fight. Influence. Generate.
The matrix is simple. A query has an intent. The AI is visible on that query, or not. You are cited, or not.
Search Engine Land analyzes it exactly like that: classify queries by intent and visibility in AI Overviews. Then decide.
Many advertisers panic. They increase bids. They cut SEO. They redo everything.
Result: they stall.
Three actions are enough.
Out of 15 sites audited this week, those who decide win. Those who hesitate lose ground.
The client applied the sorting in two days. 47 queries classified. 31 in influence mode. 12 in battle mode. 4 in generation mode.
In seven weeks, his organic clicks on cited queries grew 38%. His cost per click fell 9% on transactional queries kept in paid.
The design stayed the same. The budget stayed flat. Just a decision.
Cognitive load slows everything. The DOSE framework from Guillaume Attias, taught at BMO Academy, shows it: fewer options, more execution.
When the team saw the three columns, they stopped debating.
SEO no longer asks « how many words? ». Paid no longer asks « what budget? ». Each query has an action.
That frees the team.
One question remains. What do you do with the 47 queries that matter?
Let’s look at the matrix.
Avant de parler tactique, posons les chiffres. Sur les 47 requêtes transactionnelles suivies, l’AI Overview occupe la position zéro sur 31 d’entre elles — soit deux tiers du terrain déjà pris.
31 requêtes sur 47 captées par l’AI Overview
Répartition des requêtes transactionnelles suivies sur le cas e-commerce
The intent × AI visibility matrix
The classic mistake is treating all queries the same way.
An informational query is managed differently from a transactional query.
And above all: a query where you are cited in the AI Overview is managed differently from a query where the AI talks about your competitors.
Cross these two axes, and you get three boxes.
First box: strong transactional intent, low or nonexistent AI visibility. You fight. Paid keeps its role as a catcher of qualified clicks.
Second box: commercial or informational intent, strong AI visibility. You influence. The goal shifts from blue ranking to the citation in the answer.
Third box: broad informational intent, overwhelming AI visibility, low marginal click. You generate demand. You move the battle upstream of the search.
I’ve seen this sorting with a SaaS client. 214 queries tracked. 58% informational, 32% commercial, 10% transactional.
Before the sorting, he was bidding on 89 queries. Cost per acquisition up 31%. After the sorting, he cut 41 queries. The budget went to demand generation.
Result: 67 demos in six weeks. Before, he had 41.
AI leaves paid intact. It changes the position of the click.
Search Engine Land says it clearly: the decision to bid, influence, or generate depends on visibility in the AI Overview, beyond just search volume.
A simple table is enough:
- Transactional, you are not cited: you bid.
- Commercial, you are cited: you reinforce the citation.
- Informational, AI dominates: you create demand upstream.
You can skip the 12 columns. You can skip the 100-point scoring.
Three boxes. Three actions.
Cognitive load drops immediately.
Now let’s look at the precise mechanism of each action.
Croisez deux axes — l’intention de la requête et votre présence dans la réponse IA — et vous obtenez trois cases. Chacune appelle une action précise : fight, influence, ou generate.
L’arbre de décision : trois actions, trois cases
Intention × visibilité IA : chaque requête tombe dans une case et déclenche une action
Fight: paid remains profitable on converting queries
Fighting goes beyond doubling the budget.
Fighting means keeping paid on queries that trigger a fast conversion.
Transaction queries, those where the user wants to buy, subscribe, book an appointment, remain profitable under two conditions.
First condition: your landing page converts. I observe with my clients that pages with a unique title, a quantified proof, and a short form convert between 3 and 5%. Generic pages stay under 1%.
Second condition: the AI Overview hasn’t already diverted the intent. On « buy Salomon trail shoes size 42 », the AI answers less well. The paid click keeps its strength.
On « best trail shoe », the AI grabs attention. Bidding becomes expensive.
A case observed last month. An outdoor gear e-commerce. $12 000 in monthly paid budget.
After the appearance of AI Overviews on its commercial queries, its cost per click rose 17%. Its conversion rate stayed stable at 4.7%. But clicks fell 22%.
The math is simple. On 2 300 clicks, he made 108 sales. Now, 1 794 clicks for 84 sales.
The right move: keep paid, concentrate it.
We identified 12 high-margin transactional queries. We raised bids by 20% on those only. We cut 37 informational queries.
In three weeks, cost per acquisition fell 14%. Sales volume stayed stable.
My rule: invest in pages that convert.
A weak page leaves paid powerless.
Fighting starts with strengthening the landing page. Then bidding on a reduced perimeter.
Paid remains a lever. A lever, not a lifebuoy.
Another observation: when a page is cited in the AI Overview, its paid conversion rate also rises. The client saw it: +11% conversion rate on associated campaigns.
The battle is won upstream of the bid.
Influence: entering the answer, not the ranking
Influencing is the most profitable lever of the three.
When the AI Overview cites your brand on a commercial query, the click is almost free.
I observe this mechanism every week: a citation in the AI answer replaces ten blue positions.
An industrial client. 31 queries tracked. Zero initial citations. The AI was talking about three competitors.
We chose restructuring over producing more content.
First, we aligned each page with a precise question. We replaced « our maintenance solution » with « what is the cost of preventive maintenance on a bottling line? ».
Then, we strengthened the entities. Company profile, Organization schema, consistent mentions of the name, products, service areas.
Third move: we rewrote the answers in extractable format. 45 to 65 words. A definition. A number. A source.
Search Engine Land reminds: the AI cites what it can compress.
Six weeks later, the client was cited on 22 of the 31 queries.
Organic clicks grew 38%. Inbound calls doubled.
Side effect: paid became more effective. When a page is cited in the AI Overview, its paid conversion rate also rises.
The client saw it: +11% conversion rate on associated campaigns.
Influencing is about structure. Six weeks of work, far less than six months of production.
And cognitive load? It melts. You choose from three moves: question, entity, extract.
Aleyda Solis often says: optimization for AI is a matter of clarity.
The page must answer directly.
On the 47 queries from the first case, 31 moved to influence mode. The shift relied on structure alone. Extra content and an agency were unnecessary.
The AI receives a clear signal. The brand is authoritative on the question.
Generate demand: playing before the search
Generating demand means accepting that some searches stop generating clicks.
And that’s good news.
Broad informational queries were often low-value. The user searched « CRM comparison 2026 ». They left with a list. You paid for the click.
Now, the AI provides the list. The click disappears.
Instead of complaining, move the battle upstream.
Search Engine Land published an article the same day: « What happens before search strongly shapes the decision. » What happens before search shapes the decision.
Demand is created before the query.
A B2B client had 68% of budget on paid search. AI Overviews melted its clicks by 34%. Cost per lead exploded by 28%.
We shifted 40% of budget to demand generation.
Concretely: an expert podcast, a weekly newsletter, a YouTube video series on audit mistakes, and targeted LinkedIn posts.
Result in nine weeks: 312 inbound leads. Before, he had 187.
The lead takes longer. But it arrives better qualified.
Generated demand skips the ad click. It types the brand directly.
I observe this with my clients: branded queries increase when demand is generated upstream.
On the B2B client case, branded searches grew 27% in three months.
That’s the lever. Awareness becomes an SEO asset.
Generating demand takes time. But it’s the one asset the AI leaves untouched.
A competitor can copy your pages. Your audience remains uniquely yours.
Demand generation makes you the owner of your audience.
When AI Overviews cannibalize informational pages, your audience continues to search for you.
That’s where you win.
The DOSE framework in three questions
The DOSE framework, from Guillaume Attias at BMO Academy, applies here to the letter.
The decision: three actions. The organization: a query table. The system: a weekly sorting. The execution: two days.
The blockage comes from too many options. With 14 levers on the table, you hesitate.
Three actions reduce cognitive load.
The sequence, in order.
First question: what is the intent of the query? Transactional, commercial, informational.
Second question: is your brand cited in the AI Overview on this query? Yes. No. Sometimes.
Third question: does the marginal cost of the paid click remain profitable on this query? That’s the calculation: expected conversion rate × average order value ÷ cost per click.
If the ratio is above 3 times the cost, you fight. If you are cited, you influence. If the AI dominates and the ratio is under 2, you generate.
Skip the 50-line table.
An e-commerce client applied these three questions to 470 queries. In 48 hours.
236 queries in battle mode. 178 in influence mode. 56 in generation mode.
Paid budget was reduced by 19%. Overall revenue grew 8% in five weeks.
The most important: the CEO no longer asks the question ‘are we investing in the right place?’.
He has the answer.
Full clarity.
Marie Haynes regularly reminds: AI algorithms reward structure over volume.
I observe it in the field, every day.
Every site that applies these three actions escapes confusion within a week.
Three actions. Fourteen is overkill.
Cognitive load. That’s the bottleneck.
The SEO/SEA/generation trade-off becomes tight with paid erosion under AI Overviews. But most teams spread themselves thin tracking everything.
Sorting reduces variables. The decision becomes obvious.
And you, how many queries are you waiting for before you decide?
Live audit: your 20 queries decoded
We look together at 20 strategic queries. You see where to bid, where to influence, where to generate. You leave with a three-column plan.
Book a strategic call — 45 minFrequently Asked Questions
What is the first action to take against AI Overviews?
I sort 20 to 50 key queries by intent and visibility in the AI Overview. Three boxes are enough: fight, influence, generate.
Is paid search dead with AI Overviews?
Paid search remains profitable on purchase-intent queries, when the page converts above 4%.
How to influence an AI Overview answer?
One page, one question. Work on brand entities, then craft a 45 to 65 word extract. Dense, ready to be picked up.
When to shift paid budget to demand generation?
When the AI takes over on info queries, and the cost per click exceeds 2 times the average order value.
How long to apply this framework?
I classify 200 to 500 queries in two days. First effects appear in five to seven weeks.
